Assessing AI trustworthiness at scale
Any organisation putting AI into production ends up assessing it, whether for its own self-assessment, for conformity, or inside a regulatory sandbox. Assessment has two halves: technical tests that measure the system, and controls, compliance checklists whose questions map to specific EU AI Act provisions and return a readiness judgement rather than a metric. The AI Assessment Sandbox Configurator, developed open-source by LIST and the University of Luxembourg within the Luxembourg AI Factory, treats both as first-class catalogue entries behind one open plug-in API: qualify your AI system, select the tests and controls that fit it, and generate a sandbox à la carte that executes and reports on both, on your own infrastructure.
Alessio Buscemi is a Senior Scientist at the Luxembourg Institute of Science and Technology. His work focuses on the assessment and governance of AI systems.
I work as a Senior Research Scientist and Professor at SnT — the Interdisciplinary Research Centre on Security, Reliability and Trust (SnT) of the University of Luxembourg. My research activities currently focuses on artificial intelligence (in particular, trustworthy and secure machine learning) and I also have a strong background in software engineering. I am deputy head and one of the four permanent scientists of SnT’s SeRVal group (Security, Reasoning and Validation, 60+ members), in which I lead a team of 20+ researchers working on AI.
The key vision that drives my research is to empower Industry and Society with the capability to develop and deploy complex software solutions effectively and reliably. To achieve this, I develop and evaluate theories, methods and tools across multiple application domains and technology readiness levels.
My research is inspired from and applies to several industry partners in various domains such as finance, energy, public services, cybersecurity and space. I am deeply engaged in technology transfer and collaborative research with the industry, e.g. through the creation of spin-off companies and the leadership of private-public partnership projects at SnT.
I am leading the R&D activities of SnT in Luxembourg’s AI Factory project, a large initiative to foster AI innovation in the country. In this project, I am responsible for developing the AI Sandbox Configurator — an open-source tool that enables organisations to design and deploy customised, on-premises AI assessment sandboxes for evaluating the performance, trustworthiness, and compliance of their AI systems with EU regulations.